Homework correction method, medium, computer device and program product

Abstract

The invention discloses a homework correcting method, a medium, computer equipment and a program product. The homework correcting method comprises the following steps: acquiring a to-be-corrected target homework; the target job comprises at least one chapter; pre-generated configuration information is obtained, the configuration information comprises target chapter description information and target correction standard description information, the target chapter description information is used for describing all chapters included in the target job, and the target correction standard description information is used for describing correction standards of all the chapters; generating prompt information based on the target job, the target chapter description information and the target correction standard description information; and inputting the prompt information into a large language model to enable the large language model to correct the target homework based on the target chapter description information and the target correction standard description information.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of large language model, and particularly relates to a homework correction method, a medium, a computer device and a program product. BACKGROUND

[0002] In some teaching training scenarios, after the end of course teaching, the students need to complete course homework, and the teacher corrects the homework according to certain homework correction standards. At present, the homework correction is mainly performed in an artificial correction manner. However, in the case of a large number of homework, the artificial correction manner is low in efficiency. Moreover, the artificial correction result is greatly affected by the subjectivity of the teacher, and the correction standards of different teachers are difficult to unify. SUMMARY

[0003] In a first aspect, an embodiment of the present application provides a homework correction method, and the method comprises the following steps:

[0004] obtaining a target homework to be corrected; the target homework comprises at least one chapter;

[0005] obtaining pre-generated configuration information, wherein the configuration information comprises target chapter description information and target correction standard description information, the target chapter description information is used to describe each chapter included in the target homework, and the target correction standard description information is used to describe correction standards of the each chapter;

[0006] generating prompt information based on the target homework, the target chapter description information and the target correction standard description information;

[0007] inputting the prompt information into a large language model, so that the large language model corrects the target homework based on the target chapter description information and the target correction standard description information.

[0008] In a second aspect, an embodiment of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method described in any embodiment of the present application.

[0009] In a third aspect, an embodiment of the present application provides a computer device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the processor implements the method described in any embodiment of the present application when executing the computer program.

[0010] In a fourth aspect, an embodiment of the present application provides a computer program product, which comprises a computer program, and the computer program is executed by a processor to implement the method described in any embodiment of the present application.

[0011] In the embodiments of the present application, the chapters included in the work and the correction standards of the chapters are defined in advance through configuration information. After obtaining the target work to be corrected, the target chapter description information and the target correction standard description information in the target work are obtained based on the configuration information. Then, the prompt information is generated based on the target work, the target chapter description information and the target correction standard description information, so that the large model automatically performs work correction. On the one hand, the above process does not need manual work correction, which improves the work correction efficiency. On the other hand, the unified correction standard is defined for each chapter through the configuration information, so that different works can be corrected based on the same correction standard, which improves the uniformity of the correction standard.

[0012] It should be understood that the above general description and the following detailed description are only exemplary and explanatory, but not limiting the present application. BRIEF DESCRIPTION OF DRAWINGS

[0013] The accompanying drawings, which are incorporated into and form part of the specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0014] Figure 1 is a schematic diagram of an application scenario of an embodiment of the present application.

[0015] Figure 2 is a flowchart of a work correction method of an embodiment of the present application.

[0016] Figure 3 is a schematic diagram of a preprocessing process of an embodiment of the present application.

[0017] Figure 4 is a general flowchart of an embodiment of the present application.

[0018] Figure 5 is a schematic diagram of a correction model of an embodiment of the present application.

[0019] Figure 6 is a schematic diagram of a correction model of another embodiment of the present application.

[0020] Figure 7 is a schematic diagram of process data in a summary correction node of an embodiment of the present application.

[0021] Figure 8 is a schematic diagram of a computer device of an embodiment of the present application. DETAILED DESCRIPTION

[0022] Exemplary embodiments will be described in detail herein, with examples illustrated in the accompanying drawings. In the following description, when referring to the drawings, identical numerals in different figures represent identical or similar elements, unless otherwise indicated. The embodiments described in the following exemplary embodiments are not intended to represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.

[0023] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms, unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of.

[0024] It should be understood that although the terms first, second, third, etc. may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from each other. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0025] In order to enable people in this technical field to better understand the technical solutions in the embodiments of the present application, and to make the above-mentioned purposes, features and advantages of the embodiments of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application are further described in detail below with reference to the accompanying drawings.

[0026] In some teaching and training scenarios, after the course is over, students are required to complete course assignments, and the instructor will mark the assignments according to certain marking standards. Figure 1 The system framework of the above-mentioned teaching and training scenario is shown. The online course system may include a course delivery module and a course operation module. The course delivery module may deliver the course entrance on the online media, and users may click on the course entrance to jump to the registration page to register for the course. The course operation module includes an online course submodule and an offline grading submodule. The online course submodule is responsible for course management, student management, and the issuance of homework. The offline grading module is responsible for downloading homework, manual grading, and feedback of grading results to students. If the homework submitted by the student does not meet the passing standard, multiple rounds of homework submission, grading, and result feedback are required.

[0027] Generally, the lecturer has certain homework grading standards (referred to as grading standards) when grading homework. For example, in an actual scenario, the homework requires students to select a suitable research object and analyze the research object. The homework in some embodiments includes various chapters and grading standards as shown in the following table:

[0028]

[0029]

[0030] It can be understood that the above table is only an example. In other examples, the content of the homework and the grading standards can be different from the above example.

[0031] Currently, homework grading is mainly performed by manual grading. However, in the case of a large number of homework, the manual grading method is low in efficiency. Moreover, the manual grading result is greatly affected by the subjectivity of the lecturer, and the grading standards of different lecturers are difficult to unify.

[0032] Therefore, the embodiments of the present application propose a homework grading method. By pre-defining the chapters included in the homework and the grading standards of the chapters through configuration information, after obtaining a target homework to be graded, the target chapter description information and the target grading standard description information in the target homework are obtained based on the configuration information, and the prompt information is generated based on the target homework, the target chapter description information and the target grading standard description information, so that the large model automatically grades the homework. The above scheme does not require manual homework grading, improves the homework grading efficiency, and defines unified grading standards for each chapter through configuration information, so that different homework can be graded based on the same grading standards, improving the uniformity of the grading standards. The implementation details of the embodiments of the present application will be illustrated below with reference to the accompanying drawings.

[0033] As shown in Figure 2 the homework grading method of the embodiments of the present application includes:

[0034] Step S12: obtaining a target homework to be graded; the target homework includes at least one chapter;

[0035] Step S14: obtaining pre-generated configuration information, the configuration information including target chapter description information and target grading standard description information, the target chapter description information being used to describe each chapter included in the target homework, and the target grading standard description information being used to describe the grading standards of the chapters;

[0036] Step S16: generating prompt information based on the target homework, the target chapter description information and the target grading standard description information;

[0037] Step S18: inputting the prompt information into the large language model to make the large language model grade the target homework based on the target chapter description information and the target grading standard description information.

[0038] In step S12, the student can log in to the online course system and upload homework to the online course system. Any one or more pieces of homework uploaded to the online course system can be the target homework. The target homework can be an essay, report or thesis based on a certain argument for opinion discussion. The target homework can include at least one chapter, each chapter corresponding to a theme, and the content of each chapter can be a discussion around the theme corresponding to the chapter. Each theme can also include one or more sub-themes.

[0039] In some embodiments, the target homework can be pure text. In other embodiments, in addition to pure text, the target homework can also include tables and / or pictures. In the case of including tables and / or pictures in the target homework, the target homework can be preprocessed. The preprocessing process is described below. Figure 3 The preprocessing process is described below.

[0040] The target homework can be parsed first to determine the text, tables and pictures in the target homework. For the text in the target homework, it can be directly retained.

[0041] In the case of including tables in the target homework, preprocessing can include converting the tables into text in a specific format. After conversion, the Markdown-formatted text obtained by conversion can be inserted into the position corresponding to the table in the target homework, so as to facilitate subsequent identification of the table content during homework grading. Optionally, the text in the specific format can be Markdown-formatted text. The specific process of converting the table into Markdown-formatted text includes: using a separator (such as a vertical line “|”) to separate the contents of different cells in the same row of the table, and dividing different rows by line breaks, with a line break at the end of each row of content to start a new row. In this way, the original row structure of the table is presented, with the data of different rows displayed in sequence. In addition, the contents of the header row and the data row can be separated by a separator (such as a short horizontal line “-”), so as to distinguish between the header and the data content, making the table structure more clear.

[0042] In the case of including pictures in the target homework, preprocessing can include converting the pictures into text description information. After conversion, the text description information of the pictures can be inserted into the position corresponding to the pictures in the target homework, so as to facilitate subsequent identification of the picture content during homework grading. Specifically, the content of the picture can be identified and explained by a large language model or other picture recognition model to generate text description information.

[0043] By converting the tables and pictures in the homework into pure text, the data in the tables and the key information conveyed by the pictures can be completely preserved, so that the pure text output almost loses no content. In this way, the integrity and readability of the homework content can be ensured, providing accurate and comprehensive homework information for the subsequent homework correction process, thereby improving the accuracy and efficiency of homework correction.

[0044] In some embodiments, in the preprocessing process, a preprocessing sample can be generated based on a target homework. The pure text content in the target homework can be directly coded into the preprocessing sample; the tables in the target homework can be converted into Markdown format text and then coded into the preprocessing sample, and the position of the Markdown format text in the preprocessing sample corresponds to the position of the table in the target homework; the pictures in the target homework can be converted into text description information and then coded into the preprocessing sample, and the position of the text description information in the preprocessing sample corresponds to the position of the picture in the target homework. In the example of generating the preprocessing sample, the target homework in the subsequent steps can refer to the preprocessing sample generated based on the target homework.

[0045] In step S14, pre-generated configuration information can be obtained. The configuration information is used to configure chapter description information and correction standard description information. The chapter description information is used to describe chapters, including but not limited to the core content, main points and specific requirements of the chapters. The correction standard description information is used to describe the correction standards of each chapter, which can include the requirements of at least one dimension and the corresponding scores, including but not limited to the dimensions of content integrity, accuracy, logic, language expression, and format specification. By configuring the chapter description information and the correction standard description information, clear standards and specifications can be provided for the correction of the homework, reducing the influence of subjective factors on the correction results and improving the objectivity and accuracy of the correction results.

[0046] In some embodiments, the homework can include multiple types, and the chapters included in different types of homework can be different. For example, a homework on the current situation and development prospects of a certain industry can include chapters such as industry analysis, industry overtaking opportunities, and industry digital transformation foundation, while a homework on the application of artificial intelligence in medical image diagnosis can include chapters such as the foundation of artificial intelligence technology, the current application of artificial intelligence in medical image diagnosis, the advantages and limitations of artificial intelligence, and the prospects of the integration of artificial intelligence and medical image diagnosis. Based on this, a mapping relationship between the type of homework and the chapter description information can also be configured in the configuration information, which is used to describe the chapter description information corresponding to various types of homework.

[0047] In some embodiments, the job correction can adopt at least one job correction mode, where the job correction mode refers to a correction method with specific rules, methods and processes adopted in the process of correcting the job. Each job correction mode corresponds to a set of correction processes, and each set of correction processes includes several job correction links. Therefore, the correction process information under each job correction mode can also be configured in the configuration information, which is used to describe the job correction links included in the corresponding job correction mode.

[0048] In some embodiments, the correction mode includes a segmented correction mode and a summary correction mode. The segmented correction mode refers to a mode in which the target job is split into multiple segments, and each segment is corrected separately. The summary correction mode refers to a mode in which the target job is corrected as a whole. If the length of the target job is greater than a preset length, the segmented correction mode can be used to correct the target job; if the length of the target job is less than or equal to the preset length, the summary correction mode can be used to correct the target job. The above-mentioned preset length can be set according to experience. In actual application, the inventors found that when calling a large language model in subsequent steps, if the total length of the prompt information input to the large language model is too long, the large language model is prone to inappropriate comments and scores, and even hallucinations. The present embodiment can effectively reduce the hallucination problem of the large language model and improve the accuracy of job correction by using the segmented correction mode when the length of the target job is greater than the preset length. In the segmented correction mode, the prompt information input to the large language model each time not only includes the content of the target job itself, but also includes other information (hereinafter referred to as guiding information) for guiding the large language model to correct the job. Therefore, although the segmented correction mode can reduce the hallucination problem, it will increase the total number of characters input to the large language model. Therefore, the present embodiment uses the summary correction mode when the length of the target job is less than or equal to the preset length, so that only one guiding information needs to be input, reducing the total number of characters input to the large language model, thereby saving the computing resources.

[0049] Correspondingly, the work correction link in the segmented correction mode can include a first link of respectively correcting each chapter in the target work to obtain the correction result of each chapter, and a second link of summarizing the correction results of each chapter to obtain the correction result of the target work. By setting the first link, the correction result of each chapter in the target work can be obtained in batches in the subsequent work correction process. In the case where the length of the target work is long, the illusion problem caused by the too long input content of the large language model can be reduced, and the correction accuracy can be improved. By setting the second link, the correction results of each chapter can be summarized to accurately and comprehensively obtain the correction result of the target work as a whole. The work correction link in the summary correction mode can include a third link of correcting the target work as a whole to obtain the correction result of the target work. Since the summary correction mode corrects the target work as a whole, the correction result of the entire target work can be directly obtained. The embodiments of the present application can be applied to various application scenarios by setting multiple correction modes.

[0050] It can be understood that the above-mentioned correction modes are only exemplary descriptions, and other correction modes and correction links under corresponding correction modes can also be set according to needs in actual applications, which will not be listed one by one here.

[0051] In some embodiments, the specific content included in the prompt information can also be configured in the configuration information. For example, the prompt information can include background information, input information, output information and constraint condition information. Among them, the background information is used to describe the work correction scene and the task (in this embodiment, the work correction task) that needs to be executed by the large language model. Specifically, the background information can include work requirement information and correction standard description information. The work requirement information is used to describe the requirements that the target work needs to meet. The work requirement information can include the chapter description information in the foregoing embodiments, and can also include but is not limited to information for describing basic requirements such as the theme, type, word count, format, work submission time and manner of the work, which are used to clearly describe the core content and direction of the work. The correction standard description information is as described above, which will not be described here. The input information can include the target work as a whole or a part (such as one or more chapters in the target work) of the target work, or the correction result of each chapter in the target work. The output information is used to describe the content included in the correction result output by the large language model, such as including comments and / or scores. The constraint condition information is used to constrain the output information format and / or style of the large language model. By using the constraint condition information, the situation of the large language model outputting abstract, valueless, or even perfunctory output information can be reduced, and the effect of the output information can be improved.

[0052] Further, the prompt information applicable to different job correction links can be different. Therefore, the prompt information under each job correction link can be configured in the configuration information, including but not limited to configuring the content of the prompt information under each job correction link, and / or configuring the order of each item of the prompt information under each job correction link.

[0053] For example, when the job correction link is the first link described above, the prompt information can include, in sequence, background information, output information, constraint condition information, and input information, wherein the background information can include, in sequence, job requirement information and correction standard description information, the job requirement information and the correction standard description information can be information corresponding to one or more chapters in the target job, and the input information can include one or more chapters in the target job (the chapters included in the input information are consistent with the chapters to which the job requirement information and the correction standard description information are directed).

[0054] When the job correction link is the second link described above, the prompt information can include, in sequence, background information, summary prompt information, output information, constraint condition information, and input information. Wherein the background information can include, in sequence, job requirement information and correction standard description information, the summary prompt information is used to indicate that the current job correction link is the second link described above, for example, the summary prompt information can be: “At present, you have corrected each chapter in the target job, please summarize the correction results of each chapter below to obtain the correction result of the entire target job.” The input information can include the correction results (such as comments and scores) of each chapter in the target job, and can also include the abstracts and correction bases of each chapter. In this way, the input of the second link can be short in length, but the content is sufficient to support the output of the summary.

[0055] When the job correction link is the third link described above, the prompt information can include, in sequence, background information, correction standard description information, input information, constraint condition information, and output information. Wherein the background information can include, in sequence, job requirement information and correction standard description information, the job requirement information and the correction standard description information can be information for the entire target job, and the input information can include the entire target job. Designing the prompt information of each link in the above-mentioned manner can effectively reduce the probability of hallucination of the output result of the large language model.

[0056] In steps S16 and S18, the prompt information can be generated based on the target job and the configuration information, and the large language model is guided to correct the target job through the prompt information.

[0057] In the case that the target chapter description information and the target correction standard description information are included in the configuration information, the prompt information can be generated based on the target task, the target chapter description information and the target correction standard description information, so that the large language model corrects the target task based on the target chapter description information and the target correction standard description information.

[0058] In the case that the configuration information further includes correction process information in at least one task correction mode, the target correction mode corresponding to the target task can be determined, and the target correction process information in the target correction mode is determined, and then the prompt information in each task correction link described by the target correction process information is generated based on the target task, the target chapter description information and the target correction standard description information. Correspondingly, the prompt information of each link can be input into the large language model in turn, and the large language model can generate the correction result in the link based on the target chapter description information and the target correction standard information in the prompt information of the link.

[0059] For example, in the case that each task correction link described by the target correction process information includes the first link and the second link described above, the prompt information in the first link and the prompt information in the second link can be generated respectively. By inputting the prompt information in the first link into the large language model, the large language model can generate the correction result in the first link, i.e. the correction result of each chapter. By inputting the prompt information in the second link into the large language model, the large language model can summarize the correction results in each chapter to obtain the correction result of the entire target task.

[0060] It should be noted that in the segmented correction mode, the large language model corrects each chapter in the target task respectively and obtains the correction result of the corresponding link, so in the first link described above, the large language model needs to be called multiple times, and each time the prompt information of the first link is generated based on the corresponding chapter in the target task.

[0061] In the case that the configuration information further includes the types of the prompt information in each task correction link, the target type of the prompt information in the target task correction mode can be obtained, and in each task correction link, the prompt information of the corresponding target type is generated based on the target task, the target chapter description information and the target correction standard description information. Different types of prompt information can include different contents, and / or the order of the contents in different types of prompt information can be different. The specific contents and the order of the contents in each type of prompt information can refer to the foregoing embodiments, which will not be described here.

[0062] After generating the prompt information, the prompt information can be input into the large language model to correct the homework through the large language model to obtain a correction result. The correction result can include but is not limited to comments and / or scores. For example, in the first link, the correction result of the large language model includes comments and / or scores for each chapter; in the second link and the third link, the correction result of the large language model includes comments and / or scores for the target homework as a whole.

[0063] The correction result can be automatically uploaded to the online course system through the upload interface opened to the large language model in advance.

[0064] The general process of the embodiments of the present application will be described below. Figure 4 The general process of the embodiments of the present application will be described below. Tables are one of the common forms of homework and an important part that affects the quality (score) of homework. Tables can be converted into markdown format, thereby retaining their structure and content completely, and input into the large language model. Pictures are also one of the common forms of homework and an important part that affects the quality (score) of homework. Pictures can be converted into text description information and coded into pre-processing samples in a specific format, thereby retaining their position information and content completely, and input into the large language model. The specific format can be the reference format of Markdown.

[0065] Then, determine the correction mode and use the large language model to correct the homework. The correction mode includes segmented correction mode and summary correction mode. When the length of the target homework is less than or equal to the preset length, the summary correction mode is adopted, which is easier to debug and optimize, and can relatively simply realize the homework correction function and achieve the ideal correction effect. Moreover, the summary correction mode inputs a smaller total number of tokens to the large language model, which can save computing resources. When the length of the target homework is greater than the preset length, the segmented correction mode is adopted, otherwise, when the number of tokens input at a time is too large, the large language model will show inappropriate comments and scores, and even hallucination.

[0066] The prompt information of the large language model includes the following three types, which correspond to the first link, the second link and the third link in the foregoing embodiments respectively. The first link and the second link both use the prompt information of the “BOLI” type, that is, the prompt information includes background information, output information, constraint condition information and input information in turn. The prompt information of the above type performs well in the corresponding environment and can effectively reduce the hallucination problem of the large language model. The prompt information of the third link is better when the input information is placed before the constraint condition information and the output information because of the difference in length and output content.

[0067] Among the output information, the comments, scores, summary information of the target task, and the correction analysis content (basis for indicating the output score) can be included. By sequentially specifying the content of each item in the output information, a chain-of-thought (CoT) prompt can be formed. Specifically, before the output of the comments and scores, the summary information and the correction analysis content can be output first to guide the large language model to output comments closely related to the content of the target task and to ensure that the scores given by the large language model meet the target correction standard description information. The final output comments and scores can be stored in the label in a specified format, so as to be easily extracted in the application.

[0068] In some embodiments, a correction model can be designed, which is composed of a correction workflow definition node, an abstract chapter definition node, a summary chapter definition node, a segmented chapter definition node, a segmented node, a chapter correction node, and a summary correction node, and the specific design is as shown in Figure 5 .

[0069] The correction workflow definition node is used to define each link of the design of the task correction process. Specifically, the node can define the total prompt information template, the chapters in the task, the score label output by the large language model, the template of the prompt information of each chapter, the number and name of various task types, the comment label, and various correction standard description information.

[0070] The abstract chapter definition node is an interface for defining the comment label, score label, and title of a chapter.

[0071] The summary chapter definition node and the segmented chapter definition node are specific implementation ways of the abstract chapter definition node, wherein the summary chapter definition node is used to define the comment label, score label, and title of a chapter. In addition to defining the comment label, score label, and title of a chapter, the segmented chapter definition node can also define a segmenter of the correction standard description information and a segmenter of the target task, the segmenter of the correction standard description information is used to segment the target correction standard description information of the target task as a whole to obtain the target correction standard description information of each chapter, and the segmenter of the target task is used to segment the target task to obtain multiple chapters in the target task.

[0072] The segmented node is used to describe the characteristics of the starting position of a chapter and the characteristics of the starting position of the chapter immediately following it, so as to extract the chapter from the target task.

[0073] The chapter correction node is used to obtain the prompt information of a chapter, the chapter definition, the template of the prompt information of a chapter, the complete target task, the comment of a chapter, the content of the corresponding chapter in the target task, the output information of the large language model, the information indicating whether to segment, the correction standard description information of a chapter, the total correction standard description information of the target task, and the score.

[0074] The summary correction node is configured to store data of each chapter transmitted by the chapter correction node, including a chapter list, total prompt information, a total comment, a total prompt information template, a workflow definition of a large language model, a comment of the large language model, input information of the large language model, output information of the large language model, a correction mode and a correction standard adopted by a target job, and a score.

[0075] Since the correction workflow definition is abstracted in the present application, when a new job type exists, a new correction workflow definition is added to complete the solution access. Moreover, the summary chapter definition node can be inherited to implement summary correction, or the segmented chapter definition node can be inherited to implement segmented correction. For example, A job adopts summary correction, and B job adopts segmented correction. After access, the model is as shown in Figure 6 .

[0076] The model-driven job correction process and the instances of the chapter correction node and the summary correction node will be throughout. When the job correction process is executed and completed, the very complete process data can be saved in the database or the log by serializing the summary correction node, and the process data can be used for problem positioning or model optimization. The process data in the summary correction node includes but is not limited to comments, scores, correction modes of each node in the process, input and output of the large language model, templates of prompt information, job preprocessing samples, and the like, as shown in Figure 7 .

[0077] The present application also provides a computer device, which at least includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method of any one of the preceding embodiments when executing the program.

[0078] Figure 8 A more specific computer device hardware structure schematic diagram provided by the embodiments of the present application is shown, which can include a processor 202, a memory 204, an input / output interface 206, a communication interface 208, and a bus 210. The processor 202, the memory 204, the input / output interface 206, and the communication interface 208 are connected to each other through the bus 210 for communication within the device.

[0079] The processor 202 can be implemented in a manner of a general central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), or one or more integrated circuits, etc., for executing related programs to implement the technical solutions provided by the embodiments of the present application. The processor 202 can also include a graphics card, which can be an Nvidia titan X graphics card or a 1080Ti graphics card, etc.

[0080] The memory 204 can be implemented in a manner of a read only memory (ROM), a random access memory (RAM), a static storage device, a dynamic storage device, etc. The memory 204 can store an operating system and other application programs, and when the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 204 and called and executed by the processor 202.

[0081] The input / output interface 206 is configured to connect input / output modules to realize information input and output. The input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. The input devices can include a keyboard, a mouse, a touch screen, a microphone, various sensors, etc., and the output devices can include a display, a speaker, a vibrator, an indicator light, etc.

[0082] The communication interface 208 is configured to connect a communication module (not shown in the figure) to realize the communication interaction between the device and other devices. The communication module can realize communication through a wired manner (such as a USB, a network cable, etc.) or a wireless manner (such as a mobile network, Wi-Fi, Bluetooth, etc.).

[0083] The bus 210 includes a path for transmitting information between various components (such as the processor 202, the memory 204, the input / output interface 206, and the communication interface 208) of the device.

[0084] It should be noted that although the above device only shows the processor 202, the memory 204, the input / output interface 206, the communication interface 208, and the bus 210, in the specific implementation process, the device can also include other components necessary for normal operation. In addition, those skilled in the art can understand that the above device can also only include the components necessary to implement the solutions of the embodiments of the present application, and does not have to include all the components shown in the figure.

[0085] The embodiment of the present application provides a computer program product, comprising a computer program which, when executed by a processor, implements the method described in any of the embodiments of the present application.

[0086] The embodiment of the present application also provides a computer readable storage medium, which stores a computer program, and the program, when executed by a processor, implements the method described in any of the preceding embodiments.

[0087] The computer readable medium includes permanent and non-permanent, removable and non-removable media, and can be implemented by any method or technology to store information. The information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read only memory (ROM), electrically erasable programmable read only memory (EEPROM), flash memory or other memory technologies, compact disc read only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible by a computer device. According to the definition herein, the computer readable medium does not include transitory computer readable media, such as modulated data signals and carriers.

[0088] Each of the embodiments in the present application is described in a progressive manner, and the same and similar parts between each embodiment can be referred to each other. Each embodiment focuses on the difference from other embodiments. Especially, for the device embodiment, since it is basically similar to the method embodiment, it is described more simply, and the related parts can be referred to the part of the method embodiment. The device embodiment described above is only schematic, and the modules described as separate components can be or can not be physically separated, and the functions of each module can be implemented in the same or more software and / or hardware in the implementation of the embodiment of the present application. Part or all of the modules can be selected according to the actual needs to achieve the purpose of the embodiment of the present application. Those skilled in the art can understand and implement without creative labor.

[0089] The above is only the specific implementation of the embodiment of the present application, and it should be noted that for those skilled in the art, without departing from the principle of the embodiment of the present application, some improvements and refinements can be made, and these improvements and refinements should be regarded as the protection scope of the embodiment of the present application.

Claims

1. A method for marking homework, comprising: Get the target assignment to be corrected; The target assignment includes at least one chapter; Obtaining pre-generated configuration information, wherein the configuration information includes target chapter description information and target correction standard description information, wherein the target chapter description information is used to describe each chapter included in the target assignment, and the target correction standard description information is used to describe the correction standard for each chapter; Generate prompt information based on the target assignment, the target chapter description information, and the target grading standard description information; The prompt information is input into a large language model, so that the large language model corrects the target assignment based on the target chapter description information and the target correction standard description information.

2. The method according to claim 1, wherein the configuration information further includes a mapping relationship between job types and chapter description information; the method further includes: Obtaining the job type of the target job; Target chapter description information corresponding to the job type of the target job is determined based on the mapping relationship.

3. The method according to claim 1, wherein the configuration information further includes at least one grading process information in a homework grading mode, the grading process information being used to describe the homework grading steps included in the corresponding homework grading mode; The generating of prompt information based on the target assignment, the target chapter description information, and the target grading standard description information includes: Determining a target grading mode corresponding to the target assignment; Determine target correction process information under the target correction mode; Generate prompt information for each homework correction link described in the target correction process information based on the target homework, the target chapter description information and the target correction standard description information; Inputting the prompt information into a large language model so that the large language model corrects the target assignment based on the target chapter description information and the target correction standard description information includes: The prompt information under each homework correction link described by the target correction process information is input into the large language model in sequence, so that the large language model generates the correction result under the corresponding homework correction link based on the target chapter description information and the target correction standard description information in the prompt information under the corresponding homework correction link.

4. The method according to claim 3, wherein the at least one grading mode comprises a segmented grading mode and a summary grading mode, and the homework grading steps in the segmented grading mode include: a first step of respectively correcting each chapter in the target assignment to obtain correction results for each chapter, and a second step of summarizing the correction results for each chapter to obtain the correction results for the target assignment; The homework correction link in the summary correction mode includes a third link of correcting the target homework as a whole to obtain the correction result of the target homework.

5. The method according to claim 4, wherein if the length of the target assignment is greater than a preset length, the target correction mode is the segmented correction mode; If the length of the target assignment is less than or equal to the preset length, the target correction mode is the summary correction mode.

6. The method according to claim 3, wherein the configuration information further includes the type of prompt information for each assignment correction link; the prompt information for each assignment correction link described in the target correction process information is generated based on the target assignment, the target chapter description information, and the target correction standard description information, including: Obtain the target type of the prompt information in the target homework correction mode; Prompt information of target types under each assignment correction link described in the target correction process information is generated based on the target assignment, the target chapter description information and the target correction standard description information.

7. The method according to claim 1, further comprising: If the target job includes a table, convert the table into text information in Markdown format.

8. The method according to claim 1, further comprising: If the target job includes a picture, obtaining text description information of the picture; The text description information is added to the target job.

9. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 8 when executing the computer program.

11. A computer program product, comprising a computer program, wherein when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images